Support for mounting HubiC drive in GNU/Linux
C Shell M4 Makefile
Latest commit 342d55f Nov 27, 2016 @TurboGit Fix compilation on i686 platforms.
Fixes #135


HubicFuse is a FUSE application which provides access to Hubic's
cloud files via a mount-point.

This version contains support for DLO, symlinks and support to see other
tenant's containers. Those features are coming from:


    You'll need libcurl, fuse, libssl, and libxml2 (and probably their dev
    packages) installed to build it.

    For CentOS or similar:
        $ yum install gcc make fuse-devel curl-devel libxml2-devel \
            openssl-devel json-c-devel file-devel

        $ PKG_CONFIG_PATH=/lib64/pkgconfig ./configure

    For GNU/Debian or Ubuntu:
        $ sudo apt install gcc make curl libfuse-dev pkg-config \
            libcurl4-openssl-dev libxml2-dev libssl-dev libjson-c-dev \

        $ ./configure

    Then just compile and install:

        sudo make install


    Your Hubic Cloud configuration can be placed in a file named
    $HOME/.hubicfuse. All the following variables are required:

        client_id=[Hubic client id for the registered application]
        client_secret=[Hubic client secret for the registered application]
        refresh_token=[A refresh token you got from the script]

    Optional variables:
        get_extended_metadata=[true/false, force download of additional file metadata like atime and mtime on first directory list]
        curl_verbose=[true/false, enable verbose output for curl HTTP requests]
                curl_progress_state=[true/false, enable verbose progress output for curl HTTP requests. Used for debugging.]
        cache_statfs_timeout=[value in seconds, large timeout increases the file access speed]
        debug_level=[0 default, 1 extremely verbose for debugging purposes]
                enable_chmod=[true/false, false by default, still experimental feature]
                enable_chown=[true/false, false by default, still experimental feature]

    client_id, client_secret can be retrieved from the hubic
    web interface:

    The refresh_token can be obtained running the script provided (hubic_token)
    or with any other method you like if you follow the example at

    Then you can call hubicfuse:

        sudo hubicfuse /mnt/hubic -o noauto_cache,sync_read,allow_other

    And finaly, it can be set in /etc/fstab:

        hubicfuse /mnt/hubic fuse user,noauto 0 0

    It also inherits a number of command-line arguments and mount options from
    the Fuse framework.  The "-h" argument should provide a summary.


    Add the user into the fuse group:

       $ sudo usermod -a -G fuse [username]

    Mount using the above command without the sudo. The .hubicfuse file is
    searched into the user's home.

    To unmount use:

       $ fusermount -u [chemin]

USE with rsync:

    HubiC protocol has no support for renaming. So be sure to use the
    --inplace option to avoid a second upload for every uploaded object.


    * A segment size is limited to 5Gb (this is not hubicfuse limit, but HubiC implementation).
      So segment_above should never exceed 5Gb.
    * rename() doesn't work on directories (and probably never will).
    * When reading and writing files, it buffers them in a local temp file.
    * It keeps an in-memory cache of the directory structure, so it may not be
      usable for large file systems.  Also, files added by other applications
      will not show up until the cache expires.
    * The root directory can only contain directories, as these are mapped to
      containers in cloudfiles.
    * Directory entries are created as empty files with the content-type
    * Cloud Files limits container and object listings to 10,000 items.
      cloudfuse won't list more than that many files in a single directory.
    * File copy progress when uploading does not work, progress is shown when
      file is copied in local cache, then upload operation happens at 100%

Recent additions/fixes:
    * Support for atime, mtime, chmod, chown.
    * Large files (segmented) have correct size listed (was 0 before).
    * Multiple speed improvements, minimised the number of HTTP calls and added more caching features.
    * Fixed many segmentation faults.
    * Cached files are deleted on cache expiration when using a custom temp folder.
    * Files copied have attributes preserved.
    * Working well with rsync due to mtime support and proper copy operations.
    * Debugging support for http progress (track upload / download speed etc.)
    * Reduced traffic, skips file uploads to cloud if content does not change (using md5sum compare)
    * Major code refactoring, code documentation, extensive debugging, additional config options


    * Pascal Obry              
    * Tim Dysinger             
    * Chris Wedgwood           
    * Nick Craig-Wood          
    * Dillon Amburgey          
    * Manfred Touron           
    * David Brownlee           
    * Mike Lundy               
    * justinb                  
    * Matt Greenway            
    * Dan Cristian             

Thanks, and I hope you find it useful.

Pascal Obry